If I remember rightly, SC's design mostly keeps GC effects out of the synthesis path. SC is split into a GC'd frontend and a non-GC'd backend, and it's the bac

If I remember rightly, SC's design mostly keeps GC effects out of the synthesis path. SC is split into a GC'd frontend and a non-GC'd backend, and it's the backend that does the sound synthesis. The frontend runs the SC-language programs, whose output is a graph of sound processing components that is serialized and sent to the backend, where it updates the running synthesis graph. I think also the frontend can send new parameter values for nodes in the graph to the backend, so the frontend GC can have an effect on synthesis in cases of allocation-heavy, rapid control; but my limited experience with SC didn't involve that level of control.